StyleForge: Indoor Furniture Styling by Counterfactual Reasoning in a Hypergraph Field
2026-08-03 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors present StyleForge, a method to pick furniture pieces that look good together in a fixed room layout without changing where they are placed. It uses a special model that understands style from language descriptions and considers the whole room's furniture relationships, rather than choosing items one by one. During training, the system learns how each furniture choice fits with others, and at test time, it fine-tunes choices to avoid style clashes. Their experiments show that StyleForge creates more harmonious furniture arrangements than previous methods.
fixed-layout furniture stylingmultimodal large language modelhypergraph style fieldscene compositioncandidate distributionstyle coherenceMahalanobis energytest-time training3D-FRONT datasetfurniture retrieval
Authors
Lingwei Dang, Shishuo Shang, Pan Liu, Jiajia Cheng, Ziyan Qiu, Zhenhao Zhang, Yufei Zhu, Shenghui Huang, Qingxin Xiao, Yun Hao, Juntong Li, Qingyao Wu
Abstract
Fixed-layout indoor furniture styling requires selecting assets that form a coherent room without changing the prescribed furniture categories, positions, orientations, or scales. Existing approaches typically retrieve each asset independently or rely on static local relations, making them prone to shape, material, and color conflicts after scene composition. We introduce StyleForge, a scene-level structured selection framework built on a dynamic hypergraph style field. A frozen multimodal large language model extracts structured style priors from an open-ended style request and the fixed layout, while StyleForge maintains a learnable candidate distribution for each furniture slot. Conditioned on the target style, the dynamic hypergraph style field adaptively activates and weights layout-induced hyperedges to capture higher-order dependencies among furniture. Counterfactual style preference learning then treats each candidate as a local substitution in the current style field and evaluates its contextual compatibility using Mahalanobis energies. Training alternates between optimizing the style field and the candidate logits. At inference, the model remains frozen and test-time training updates only room-specific candidate logits, progressively correcting cross-slot style conflicts as the global scene context evolves. Experiments on 3D-FRONT demonstrate state-of-the-art furniture retrieval and scene-level style coherence, producing more coherent fixed-layout furniture arrangements than object- and scene-level retrieval baselines.